Skip to content

Brainchip-Inc/models

Repository files navigation

Akida Model Zoo

Introduction

The Akida Model Zoo expands our foundation models with a hand-picked collection of models accelerated by the Akida 2.0 IP. Designed for developers, researchers, and AI enthusiasts, these ready-to-use models make it easier than ever to explore, build, and innovate with the Akida solution.

Models

Both float and quantized models are available, with quantized versions converted and evaluated on the Akida solution. For each model, the number of nodes required to run on a minimal Akida IP configuration is provided, enabling straightforward assessment of performance and deployment needs.

In addition, some models can be evaluated directly through Akida Cloud ☁️, offering a convenient way to explore and experiment without local hardware.

Domain Use case Architecture Resolution Dataset #Params Quantization Accuracy F1 Score MSE Minimal #Nodes
Vision Classification MobileNetV1_1.0 128 CIFAR-10 2.25M 8 91.92% 5 ☁️
Vision Classification MobileNetV1_1.0 224 Oxford_Flower 3.3M 8 91.08% 7
Vision Classification MobileNetV1 0.5 224 SIIM-ISIC 3.14M 8 98.16% 86.14% 4 ☁️
Vision Classification MobileNetV1 0.5 224 ODIR-5K 3.15M 8 92.00% 97.83% 4 ☁️
Vision Classification MobileNetV1 0.5 224 ECG 3.14M 8 83.27% 89.18% 4 ☁️
Vision Classification MobileNetV1 0.5 224 Retina OCT 3.15M 8 93.30% 98.66% 4 ☁️
Vision Classification MobileNetV2 1.0 224 ImageNet 3.5M 8 70.35% 7
Vision Classification MobileNetV2 0.75 160 ImageNet 2.6M 8 62.85% 4 ☁️
Vision Classification MobileNetV2 0.35 96 ImageNet 1.2M 8 43.47% 2 ☁️
Vision Classification MobileNetV4 1.0 224 ImageNet 3.77M 8 71.86% 8
Vision Classification MobileNetV2_1.0 128 CIFAR-10 2.25M 8 93.96% 5 ☁️
Vision Classification MobileNetV2_1.0 224 Oxford_Flower 2.4M 8 91.97% 7
Vision Classification MobileNetV4_1.0 128 CIFAR-10 2.5M 8 94.72% 7
Vision Classification MobileNetV4_1.0 224 Oxford_Flower 2.6M 8 85.41% 8
Vision Classification spatiotemporal 224 FallVision 1.34M 8 98.36% 16
Vision Classification MLP 784 MNIST 203.5K 8 98.05% 1 ☁️
Vision Classification LogisticRegression 784 MNIST 7.9K 8 84.52% 1 ☁️
ECG Classification 1DCNN 360 MIT-BIH 74K 8 97.3% 1 ☁️
ECG Anomaly Detection 1DCNN 144 ECG5000 290K 8 94.0% 1 ☁️
Tabular Classification LogisticR. 30 Breast_Cancer 169 8 93.9% 1 ☁️
Synthetic Regression MLP 1 1D_Curve 6.2K 8 0.136 1 ☁️
Vision Detection AkidaNet18/CenterNet 224 Soda_bottle 2.43M 8 91.53% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Trail_camera 2.43M 8 84.74% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Road_signs 2.43M 8 65.46% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Furniture 2.43M 8 79.21% 6 ☁️
Vision Detection AkidaNet18/CenterNet 384 Aerial_Cows 2.43M 8 30.91% 16
Vision Detection AkidaNet18/CenterNet 224 Bees 2.43M 8 59.99% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Cable_Damage 2.43M 8 77.32% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Corrosion 2.43M 8 39.06% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Digits 2.43M 8 92.32% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Face_Detection 2.43M 8 75.98% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Hand_Gestures 2.43M 8 53.52% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 License_Plate 2.43M 8 96.22% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Liver_Disease 2.43M 8 40.57% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Phages 2.43M 8 67.18% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Sign_Language 2.43M 8 85.88% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Stomata_Cells 2.43M 8 52.14% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Underwater_Objects 2.43M 8 44.89% 6 ☁️
Vision Detection AkidaNet18/CenterNet 384 Ships_Detection 2.43M 8 39.60% 12
Vision Detection AkidaNet18/CenterNet 224 Bone_Fracture 2.43M 8 60.70% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Facial_Expression 2.43M 8 75.40% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Pothole_Detection 2.43M 8 57.20% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Search_And_Rescue 2.43M 8 77.00% 6 ☁️
Vision Detection AkidaNet18/CenterNet 224 Traffic_Detection 2.43M 8 71.80% 6 ☁️
Vision Detection AkidaNet18/CenterNet 384 Weed_Crop 2.43M 8 47.70% 16

Download

Git Clone

To avoid downloading the models during cloning due to their large size:

GIT_LFS_SKIP_SMUDGE=1 git clone [email protected]:Brainchip-Inc/models.git

To download a specific model:

git lfs pull --include="[path to model]" --exclude=""

To download all models:

git lfs pull --include="*" --exclude=""

GitHub UI

Alternatively, you can download models directly from GitHub. Navigate to the model's page and click the "Download" button on the top right corner.

Model Visualization

For a graphical representation of each model's architecture, we recommend using Netron.

About

An extension of Akida model zoo with pre-trained models

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Contributors

Languages